Probing the Etic Vs. Emic Nature of Consumer Ethnocentrism Srinivas Durvasula Marquette University, [email protected]

Probing the Etic Vs. Emic Nature of Consumer Ethnocentrism Srinivas Durvasula Marquette University, Srinivas.Durvasula@Marquette.Edu

Marquette University e-Publications@Marquette

Marketing Faculty Research and Publications Marketing, Department of

1-1-2014 Probing the Etic vs. Emic Nature of Consumer Ethnocentrism Srinivas Durvasula Marquette University, [email protected]

Steven Lysonski Marquette University, [email protected]

Srinivas Durvasula (USA), Steven Lysonski (USA) Probing the etic vs. emic nature of consumer ethnocentrism: cross-national evidence Abstract The consumer ethnocentrism concept and its measure, the CETSCALE, remain very popular in cross-national research chiefly because they serve as a means to understand consumer attitudes toward imports. But the usage of consumer ethnocentrism and its measure are based on the premise that they have universal or etic properties. Conflicting studies, however, find that the scale’s structure is far more complicated than initially believed, and that it may not be uni- dimensional as originally proposed. Is it possible that the consumer ethnocentrism concept and its measure are culture bound? The goal of this study is to resolve this ambiguity. Introduction One such construct that was developed in North America and applied frequently in other countries is The field of consumer behavior is heavily dependent the consumer ethnocentrism concept and its on measurement scales to quantify psychological corresponding measurement scale (CETSCALE). We characteristics of consumers. As globalization use the consumer ethnocentrism construct as an accelerates, researchers are keen to employ these exemplar to demonstrate how to probe the emic vs. measures for cross-national or multi-country research. etic nature of a concept and its measure. Etic deals Of concern, however, is that Western countries have with the universality of the consumer ethnocentrism been the primary source of these scales. Watters concept and the psychometric applicability of the (2013) discusses a 2008 survey of the top psychology CETSCALE to other countries; in contrast, the notion journals (often a source for consumer research) that of emic focuses on the concept or scale being relevant shows that 96% of the subjects used in psychological only to one culture. Since the CETSCALE has been studies from 2003 to 2007 were Westerners; 70% were administered to foreign consumers with an implied from the USA. The goal of this paper is to explore the confidence a priori that it is as valid and reliable as it is robustness of the consumer ethnocentrism concept in the USA, researchers have accepted the scale with a and its measure regarding their application in level of blind confidence (i.e., imposed etic) in its diverse countries. efficacy to capture the ethnocentrism property. More One issue that begs an answer deals with the etic vs. attention needs to be given to etic considerations emic question concerning the relevance of these regarding the concept per se and its scale. scales to foreign countries, particularly non-Western The scale is also assumed to mirror ethnocentrism as ones. Conceptually, an etic construct is a theoretical a uni-dimensional construct which is again an idea that is assumed to apply to all nations or imposed etic. But is this mirror more complicated? cultures while an emic construct is one that applies Are there more dimensions in the scale than to only one nation or culture. Recent research has believed and are these dimensions specific to found that aspects of human nature that are different cultures? Douglas and Nijssen (2003) considered universal (i.e., etic) may in fact be caution researchers that the scale may not be uni- culturally bound and a product of social learning in dimensional as they found for the Netherlands. that culture (i.e., emic) (Henrich and Boyd, 1998; Hence, the faith that researchers have taken in the Henrich, Heine and Norenzayan, 2010). Indeed, structure of the scale and its imposed etic may be one’s culture may deeply shape human cognition overly optimistic. One foundation of the CETSCALE and influence consciousness and decision making. is its morality dimension; yet, Henrich, Heine and Cultural bias arises when researchers assume that an Norenzayan (2010) observed that the moral reasoning emic construct is intrinsically etic. The result is found in Western societies is different elsewhere. referred to as an imposed etic where a culture specific construct is incorrectly imposed on a Our research investigates the etic and emic different culture. In fact, a number of scholars have properties of the consumer ethnocentrism construct argued that some scales developed in the USA may and the CETSCALE, and examines if the scale is be irrelevant to foreign consumers (de Mooij, 2010; uni-dimensional. Specifically, is the consumer Douglas and Nijssen, 2003; Herche, Swenson, and ethnocentrism construct equivalent cross-nationally Verbeke, 1996). In essence the argument deals with or is it country specific? Correspondingly, is the the etic vs. emic debate. CETSCALE invariant across cultures or should researchers be chided for ignoring the emic question as they confidently use the scale beyond North Srinivas Durvasula, Steven Lysonski, 2014. America? To achieve our goal, we choose two

7 Innovative Marketing, Volume 10, Issue 1, 2014 culturally different countries for our examination: superiority of one’s own group, Douglas and Nijssen Singapore and New Zealand. (2003) raised questions about the construct’s relevance to smaller countries that have open The rest of the paper is organized as follows. First, economies, high levels of foreign trade, few major we discuss the origin and application of the domestic manufacturers and dependence on imported ethnocentrism concept and its measure. Sub- products. Their concern, therefore, is rooted in the sequently, we examine emic and etic issues relevant emic theme. Hence, we do not know whether to this construct and scale followed by research consumer ethnocentrism and the CETSCALE can be questions and our methodology. The results are then applied reliably in such countries. Our study discussed with conclusions and implications. addresses this shortcoming by comparing the findings 1. Background discussion from the city-state of Singapore, which is at the hub of South East Asia, with those from New Zealand, The concept of ethnocentrism was introduced in sociology by Sumner (1906) to distinguish between which is relatively more isolated, but a developed in-groups with which a person identifies and out- economy nonetheless. groups lacking this identification. Ethnocentric 2. Emic vs. etic issues of the consumer people prefer the in-group over the out-groups to ethnocentrism concept such an extent that symbols and values of the in- group become an object of pride whereas symbols Cross-culturally, a concept or its measure can be and values of the out-group are likely to become either emic or etic. Emic models view a specific objects of contempt (LeVine and Campbell, 1972). behavior as specific to that culture; hence, consumer Shimp and Sharma (1987) extended this concept to behavior must be understood in the context of a marketing and called it “consumer ethnocentrism,” particular culture. In contrast, etic models view a defined as “the beliefs held by American consumers specific behavior as universally generalizable, about the appropriateness, indeed morality, of allowing for comparisons among consumers in purchasing foreign made products”. Based on this varying cultures (i.e., cross-cultural) on that definition, purchasing foreign products is considered behavior. Hence, if it can be shown that the wrong because it hurts the domestic economy, consumer ethnocentrism concept can be concep- causes loss of jobs, and is unpatriotic (Shimp and tually understood similarly by respondents and that Sharma, 1987). To highly ethnocentric consumers, its measure is equivalent (or invariant) across domestic products are viewed as superior while cultures, we have established some degree of etic foreign made products are objects of contempt. proof of its universal application. Of the many Accordingly, Shimp and Sharma (1987) developed issues regarding cross-cultural research methodo- the consumer ethnocentric tendency scale, logies, the equivalence of concepts and their abbreviated as the CETSCALE with 17 items and a measures is viewed as the single dominant one (cf., reduced 10-item version. Albaum and Baker, 2005; Berry, 1980; Craig and Empirically, studies show that consumer ethno- Douglas, 2005). centrism negatively impacts consumer behavior. For an imposed etic validity to be acceptable, there Specifically, consumer ethnocentrism related unfavo- must be equivalence of consumer ethnocentrism and rably to the following: attitudes toward the ad its measure between the source nation of the (Reardon, Miller, Vida and Kim, 2005), purchase construct and the country where it is to be applied intention of foreign brands (Suphellen and Ritten- (Albaum and Baker, 2005). If foreign consumers burg, 2001), attitudes toward imports (Sharma, construe a construct differently or respond in unique Shimp and Shin, 1995), evaluations of foreign ways to its scales, Douglas and Craig (1983) warn services (de Ruyter, Van Birgelen and Wetzels, us that “relevant constructs will be unique to a given 1998), preferences toward foreign products over country” and therefore lack this universal quality. If domestic products (Klein, Ettenson and Krishnan, the concept can be universally understood, we can 2006), and attitudes toward outsourcing (Durvasula conclude that it is “culture free” and proceed to and Lysonski, 2009). Unless etic and emic issues are apply it in a pan-cultural sense (Craig and Douglas, examined, we do not know if the observed mean 2005; Herche, Swenson and Verbeke, 1996). differences in these studies on the construct’s Sekaran (1983) cautions that unless we have measure reflect true cross-national differences or established the etic characteristics of the construct, measurement artifacts. we face a “pseudoetic” (or imposed etic) dilemma in Most cross-national studies using the CETSCALE using the scale to make cross-cultural comparisons. have focused on large economically developed We expect that the consumer ethnocentrism concept countries with significant domestic and foreign has a universal understanding given its underlying competition. Because consumer ethnocentrism is theory discussed above. In this regard, we followed associated with feelings of nationalism and the the approach of Herche, Swenson and Verbeke 8 Innovative Marketing, Volume 10, Issue 1, 2014

(1996) by using qualitative field studies; these were prescribed a number of different hierarchically conducted in Singapore and New Zealand to linked equivalence tests to establish measurement determine if the consumer ethnocentric construct equivalence. Each successive test in the hierarchy existed in consumers’ minds. We discussed the assumes increasingly stronger measurement consumer ethnocentrism concept with over 300 equivalence across cultures as discussed below. consumers in their early 20s in university class settings to understand their conceptualization (or 3.1. Structural invariance. Also known as universality of the concept). In both countries, the construct equivalence or configural invariance, this authors did not discern any confusion in grasping form of equivalence tests whether the set of scale the consumer ethnocentrism concept and its items has the same pattern (structure or relevance to their country. configuration) of factor loadings with the construct to be measured across cultures. For Singapore, the government constantly promotes themes about “pride in being a Singaporean” which 3.2. Metric invariance. This second test in the the young consumers recalled vividly, allowing them hierarchy assumes structural invariance and invariant to comprehend the ethnocentrism idea effectively. Of relationships between observed indicators and the special note, in New Zealand a “Buy New Zealand latent concept (i.e., factor loadings) across cultures. Made” campaign started in 1988 made the Also referred to as measurement unit equivalence, it ethnocentrism concept and the obligation to buy implies that across-cultural groups there is equality of domestically made goods palpable. To quote the web the measurement units or scale intervals. Metric site concerning this campaign “When you buy a New invariance is necessary for the comparison of Zealand produced product or service, you’re helping to difference (i.e., mean-centered) scores across cultures create jobs, promote growth. As every cricketer (Steenkamp and Baumgartner, 1998). knows, a run saved equals a run scored – so you’re In CFA (confirmatory factor analysis), metric giving our country a double whammy benefit when invariance can be established by showing no you buy New Zealand rather than from another significant drop in fit between the metric invariance country. We can be proud of the quality of our model and the structural invariance model. In EFA, products” (http://www.buynz.org.nz/MainMenu). In invariance of factor loadings can be established on sum, comparisons of the responses amongst the two the basis of the size of Tucker’s phi index (or sets of young consumers from both countries, who congruency coefficient) (Tucker, 1951). For any were similar in age and educational background, factor, this index is a measure of the degree of showed understanding of the concept and its agreement between the factor loadings of items measurement. Hence, we view our qualitative result at from two different cultures. When cross-cultural the conceptual level as evidence of the etic property; comparisons involve more than two countries, the concept was clearly not emic. Tucker’s phi index is computed for each factor and Given that we established the etic quality of for each pair of countries. The formula for Tucker’s consumer ethnocentrism concept qualitatively, the phi index is as follows: next step was to determine whether the CETSCALE xy also had the etic property regarding its purported ¦ ii )(,xy ) i , universal uni-dimensional nature and defined factor 22 x x structure. We now probe further into the etic vs. ¦¦ii emic question about the CETSCALE using a series where xi and yi are the loadings of variable i on factor x of tests for measurement equivalence. Herche, and y, respectively, i = 1, . . ., n. The Tucker’s phi Swenson and Verbeke (1996) used a similar index is not sensitive to scalar multiplication of x and sequence of analyses to establish the etic qualities of y, implying that it measures factor similarity the scales they examined. independently of the absolute size of the loadings. 3. Assessing the emic vs. etic nature of the Values of phi higher than 0.95 are recommended for CETSCALE assuming metric invariance (Van de Vijver and Poortinga, 1994). Cross-nationally, a concept such as consumer ethnocentrism may be etic but its measurement scale 3.3. Scalar invariance and item bias. As argued by such as the CETSCALE may not be. In such a Steenkamp and Baumgartner (1998), even after scenario, we cannot use the measure in comparative establishing metric invariance, scores on the latent research. To certify that a scale has an etic quality, it variable can still be uniformly biased upward or is essential to establish that it has measurement downward, when the origin of the scale is not the equivalence (i.e., it is cross-nationally equivalent). same across cultures. It means that people who have Steenkamp and Baumgartner (1998) and others the same level of a latent trait (but are from different

9 Innovative Marketing, Volume 10, Issue 1, 2014 cultures) exhibit higher or lower scores on the bias (called uniform and non-uniform) that affect measure. To perform mean comparisons across cross-national mean comparisons (see Figure 1). cultures, it is also necessary that the origins of the To detect item bias, Van de Vijver and Leung (2011) scale items (i.e., intercepts) are the same across recommend the following procedure for measures those cultures (i.e., scalar invariance). In CFA, using interval scaled (e.g., 7-point rating scale) items scalar invariance can be established when there is administered in two cultures (A and B). First, for a comparable fit between the scalar invariance model set of items that represent the same dimension of the (invariant loadings and intercepts) and the metric concept, the composite score across all those scale invariance model (invariant loadings but not items is computed for each respondent. Then, the invariant intercepts). entire sample from cultures A and B is split Scalar invariance is closely related to the concept of according to the composite score – based on item bias or differential item functioning (DIF) (Van percentile or quartile scores. The number of score de Vijver and Leung, 2011). Presence of scalar levels is determined arbitrarily, and they may range invariance implies that items do not exhibit bias from “very low” to “very high”. The number of cross-culturally. An item exhibits bias if participants for any score level should be neither too respondents with the same level of latent trait (e.g., big nor too small; the recommended sample size for they are equally consumer ethnocentric) do not have any score level is 50 (Van de Vijver and Leung, the same mean score on the item across cultures 2011). Next, an analysis of variance is performed for (Van de Vijver and Leung, 1997), and the likely each scale item separately. In this analysis, score reasons include poor item translation, ambiguities in level and culture are treated as the independent the original item, and inappropriateness or low variables and item score is the dependent variable. familiarity of the item in certain cultures. Van de The mean item scores of the respondents in the two Vijver and Leung (2011) show two types of item cultures are plotted against the score level.

Fig. 1. Examples of biased and unbiased items As shown in Figure 1, the items are unbiased if the show either uniform or non-uniform bias (or both), curves for the two cultures are close to each other; then it is futile to make cross-cultural comparisons in ANOVA, this means there is only a significant based on composite scale scores. In this context, the effect for score level. A uniform bias means that the question that researchers have to address is whether item mean score is systematically higher or lower it makes sense (theoretically and practically) to for one culture as compared to the other. In delete those items from the scale that exhibit bias ANOVA, this bias can be detected when there is a cross-nationally. In sum, to establish the etic nature significant main effect for culture. A non-uniform of the scale and to perform cross-national mean bias implies that the mean item score varies for comparisons, it is imperative to show that the scale various score levels, as evidenced by a significant possesses similar dimensionality, high reliability, culture by score level interaction. When the items cross-national measurement equivalence – structu-

10 Innovative Marketing, Volume 10, Issue 1, 2014 ral, metric, and scalar – and the scale items do not RQ2: Does the CETSCALE have structural exhibit any bias. equivalence for both Singapore and New Zealand? In the development of CETSCALE, Shimp and RQ3: Does the CETSCALE have metric (or measure- Sharma (1987) found the scale to have uni-dimen- ment unit) equivalence for both Singapore and New sional factor structure and high reliability. Netemeyer, Zealand? Durvasula and Lichtenstein (1991) established the RQ4: Does the CETSCALE have scalar equivalence scale’s cross-national applicability based on data for both Singapore and New Zealand? collected from larger nations: Germany, Japan, France, and the USA. They used an imposed etic in viewing RQ5: Do individual items of the CETSCALE exhibit the construct and scale as universal, but their study did non-differential functioning for both Singapore and show the scale’s factor structure to be similar to the New Zealand? one conceptualized by Shimp and Sharma (1987). Subsequently, other studies validated the uni- 5. Method dimensional nature of the scale in Russia (Durvasula, Data was collected in New Zealand and Singapore. Andrews, and Netemeyer, 1997), Korea (Sharma, Both countries are economically developed, have a Shimp and Shin, 1995), Azerbaijan (Kaynak and Kara, large middle class population with considerable 1996), and Spain (Luque-Martinez, Ibanez-Zapata, and purchasing power, and host competing multinational del Barrio-Garcia, 2000). Contrarily, other studies corporations. Differences do exist between them. have begun to question the scale’s dimensionality Singapore, a small city-state, depends heavily on (Supphellen and Rittenburg, 2001; Vida and Damjan, foreign trade since it has few domestic manufacturers 2001). For example, Douglas and Nijssen (2003) of consumer goods. New Zealand’s inhabitants are found the scale to have two dimensions while also largely Anglo unlike Singapore’s multi-cultural Marcoux, Filiatrault, and Cheron (1997) found the population. As such, Singaporean consumers are less scale to have three. Mavando and Tan (1999) even likely to be threatened by imports resulting in suggested that consumer ethnocentrism represents a different ethnocentric tendencies compared to New higher-order dimension consisting of three first-order Zealanders. The choice of these two countries allows dimensions, which they labeled as morality, economic investigation of Douglas and Nijssen’s (2003) caveats rationality, and economic animosity. Such divergent regarding consumer ethnocentrism’s universality. findings are problematic concerning etic assumptions of the scale. Given that there is no conclusive evidence The sample consisted of 127 young consumers in New one way or the other about the etic nature of the Zealand and 145 in Singapore. To make cross-national CETSCALE, we re-examine its cross-cultural comparisons possible, we matched sample demo- applicability. We propose the following research graphics in the two countries in terms of educational questions regarding the etic quality of the background, age, and gender composition. The survey CETSCALE. Failure to validate each question (written in English) consisted of the 17-item presents a red flag concerning the use of the scale in CETSCALE and other validation measures. Table 1 other countries. shows the alternative scale versions that we analyzed. Responses to individual scale items ranged from 1 4. Research questions (strongly disagree) to 7 (strongly agree). The results RQ1: As proposed by Shimp and Sharma (1987), is below are presented in a hierarchical order starting the CETSCALE uni-dimensional with high reli- with the most rudimentary analyses proceeding to the abilities for both Singapore and New Zealand? most complex ones. Table 1. Factor models of CETSCALE examined in this study

Model Items (selected from the original scale as shown above) 17 – Item scale 1-factor model 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 (Shimp and Sharma, 1987) Hierarchical (2nd order) model (Mavado & Tan, 1999) Factor 1 5, 6, 11, 14, 17 (label: Morality) Factor 2 1, 2, 3, 4, 7, 8, 9 (label: Economic rationality) Factor 3 10, 12, 13, 15, 16 (label: Economic animosity) 2nd order factor Related to the three first order factors Bi-factor model 3 first order factors as in Mavado & Tan (1999) and one general factor for all items 10-item reduced scale 1-factor model 2, 4, 5, 6, 7, 8, 11, 13, 16, 17 (Shimp and Sharma, 1987)

11 Innovative Marketing, Volume 10, Issue 1, 2014

Table 1 (cont.). Factor models of CETSCALE examined in this study

Model Items (selected from the original scale as shown above) 2-Factor model (Douglas & Nijssen 2003) Factor 1 4, 5, 6, 7, 8, 11, 13, 17 (general ethnocentrism) Factor 2 2, 16 (nuanced attitude towards imports) Bi-factor model 2-factor model as above with one general factor for all items 6-item reduced scale 1-factor model 4, 5, 6, 7, 8, 11 (Douglas & Nijssen, 1999) 1-factor model 2, 4, 7, 11, 13, 17 (Klein, Ettenson & Krishnan, 06)

Note: Shimp and Sharma (1987) provide the complete description of the 17-item CETSCALE; we presented the scale in the same order as they appeared in the original study. 6. Results technique as a “second generation” one that is superior to “first generation” techniques such as EFA. Table 2 6.1. National level EFA of 17-item scale. Analysis presents results of CFA analysis of the one-factor of the 17-item CETSCALE revealed three factors in model. The model fit is deemed reasonably good if Singapore and two in New Zealand with the first SRMR (standardized root-mean-square residual) is factor explaining over 50% of the variance for both close to or below .08 and CFI (comparative fit index), samples. All items exhibited high loadings (above TLI (Tucker-Lewis Index), and IFI (incremental fit 0.4) on the first factor while showing inconsistent index) are close to or above 0.95 (Brown, 2006). For loadings on the remaining factors. Hence, the scale acceptable model fit, CFI and TLI values must be appears to be uni-dimensional. between 0.90 and 0.95 (Bentler, 1990). Based on these 6.2. National CFA of the 17-item scale. To establish yardsticks, the one-factor model exhibits a reasonably additional support for scale dimensionality, we good fit for both nations. All items also have performed CFA; Bagozzi and Yi (2012) described this reasonably high factor loadings (above 0.4). Table 2. Analysis of the 17-item scale

Model Fit statistic New Zealand Singapore F2 (df) 317.15 (119) 274.31 (119) CFI 0.95 0.95 TLI 0.95 0.94 1-factor model IFI 0.95 0.95 SRMR 0.07 0.07 Item Loadings 0.52-0.83 0.50-0.85 F2 (df) 285.07 (116) 234.46 (116) CFI 0.96 0.96 TLI 0.95 0.96 IFI 0.95 0.96 Second order model SRMR 0.07 0.07 (Mavado & Tan, 1999) Item loadings (Fac 1) 0.59-0.84 0.57-0.83 Item loadings (Fac 2) 0.59-0.84 0.53-0.84 Item loadings (Fac 3) 0.51-0.82 0.61-0.85

J1 11

J2 0.92 0.88

J3 0.95 0.95 F2 (df) 227.10 (104) 195.80 (104) CFI 0.97 0.97 TLI 0.96 0.96 IFI 0.96 0.97 Bi-factor model SRMR 0.06 0.06 Item loadings (Fac 1) 0.04-0.39 0.05-0.62 Item loadings (Fac 2) 0.11-0.55 0.05-0.56 Item loadings (Fac 3) 0.03-0.55 0.09-0.56 Item loadings (Gen Fac) 0.48-0.84 0.54-0.87

12 Innovative Marketing, Volume 10, Issue 1, 2014

Does the hierarchical three factor model suggested Conclusion to RQ2: The CETSCALE has structural by Mavado and Tan (1999) offer a better fit to the equivalence for both Singapore and New Zealand. 2 data? The F difference test reveals a better fit (vs. Next, we compared the structural invariance model the one-factor model). However, given the to the metric invariance model. While the sample 2 sensitivity of F to sample size, Marsh (1994) size sensitive chi-square difference test shows suggested that alternative fit indices (e.g., CFI, TLI, significant fit difference (F2 difference = 37.77, 16 SRMR) also be considered when comparing fit of df., p < 0.05), the values of CFI (0.95), TLI (0.95), alternative models. Table 2 shows that CFI and TLI and IFI (0.95) (which are the other fit indices for the hierarchical model are only marginally better recommended by researchers in such scenarios) are than those for the one-factor model. SRMR remains stable and acceptable. Also, the Tucker’s Phi index unchanged. The path estimates from the second- for the one-factor model of 0.995 is well above the order factor to the three first-order factors (J1, J2, J3) recommended level of 0.95 for invariance of factor ҏin Table 2 are very high and close to one. loadings between the two samples. Collectively, these results mean that the three first Conclusion to RQ3: The CETSCALE does have order factors in the hierarchical model are highly metric (or measurement unit) equivalence for both correlated and indistinguishable from one another. Singapore and New Zealand. Hence, the original one-factor model has better support than the alternative hierarchical model. The In our next analysis, we compared the fit of the one-factor model is also more parsimonious and metric invariance model with the scalar invariance easier to interpret. model (equal loadings and intercepts). The fit indices of this model (CFI, TLI and IFI) are all 0.95 Next, we examined a bi-factor model. The bi-factor and remained unchanged vs. those of the metric model retains the three first-order factors but adds a invariance model. general factor that is related to all 17 items. Similar to the hierarchical factor model, this model too Conclusion to RQ4: The CETSCALE does have provides a better fit (vs. 1-factor model) based on scalar equivalence for both Singapore and New the F2 difference test, but, like the hierarchical Zealand. model, it too is indistinguishable from the 1-factor Item analysis is the last step. We followed the model based on CFI, TLI, and SRMR. In short, the procedure as described earlier in this manuscript and one-factor 17-item scale provides a reasonably good divided all respondents, irrespective of their fit and is preferred over other configurations country, into “high” (top 1/3 percentile), “medium” because of its parsimony. (middle 1/3 percentile), and “low” (bottom 1/3 6.3. National level CFA of the 10-item and 6-item percentile) score level groups based on each scales. Shimp and Sharma (1987) developed a subject’s composite CETSCALE score. reduced version of the CETSCALE with 10 items With score level and country as the independent versus 17 for the full scale. Subsequently, others variables and item as the dependent variable, we proposed shorter versions with 6 items (Nijssen, then performed 17 different analyses of variance, Douglas and Bressers, 1999; Klein, Ettenson and with one ANOVA per each scale item. The results Krishnan, 2006). CFA analyses support the fit of the of particular interest are the main effect of country uni-dimensional model over its rivals for both the and the interaction between country and score level. 10-item and 6-item scales. The composite reliability The main effect is significant for scale items 1 and indices are also high in both New Zealand and 3, suggesting a possible uniform item bias for those Singapore. For the sake of brevity those analyses are two items only. The interaction effect is significant not presented in this paper. only for scale item 5, suggesting a possible non- uniform item bias. However, van de Vijver, Conclusion to RQ1: The CETSCALE is uni-dimen- Valchev, and Suanet (2009) suggest that an item is sional and possesses high reliabilities. biased only if the proportion of variance accounted The results presented below examine the measure- for by the main effect of country and the interaction ment equivalence of the 17-item uni-dimensional effect of score level and country is at least 0.06. In CETSCALE based on multiple group CFA. our case, the effect size estimates are less than 0.06. Therfore, we conclude that there is no differential 6.4. Testing measurement equivalence of the 17- item functioning for any of the CETSCALE items. item scale. The high (above 0.4) and significant factor loadings and acceptable fit indices (CFI = Conclusion for RQ5: Individual items of the 0.95, TLI = 0.95, IFI = 0.95) support invariant factor CETSCALE do exhibit non-differential functioning structure in the two samples. for Singapore and New Zealand.

13 Innovative Marketing, Volume 10, Issue 1, 2014

Discussion and implications North America as it would be if it were an emic concept. Instead, it has an etic quality that allows it be The inexorable pace of globalization has led to used in cross-cultural studies with confidence. significant interest in cross-national research, particularly in the field of consumer behavior. The Our findings offer several implications for researchers typical research practice is to borrow measurement exploring consumer ethnocentrism in other cultures. scales from countries (generally the USA) where the Our central argument is that cross-cultural research construct was developed and apply them cross- needs to examine the emic and etic qualities of a nationally to detect consumer differences. Unfortu- construct and scale. This admonition is given strong nately, use of these “borrowed” scales may be flawed credence by Henrich, Heine and Norenzayan (2010) without previous rigorous cross-national tests. These who urge researchers to establish the construct’s tests must examine etic issues looking at the concept’s meaning in another culture before concluding that it is universal understanding and emic issues relating to the universal to human behavior. Several theorists have concept being understood only by one culture. asserted that neither the etic nor emic perspectives are Moreover, the measure of the concept must exhibit sufficient when used singularly (Segall et al., 1990; cross-culturally invariant psychometric properties. Herche, Swenson and Verbeke, 1996); instead, both must be investigated. Transferring instruments from For cross-national use of the CETSCALE measure, one culture to another without exploring these the lack of consensus on scale dimensionality hinders properties may produce spurious cross-cultural its psychometric credibility in making mean conclusions. If cultures are deemed to be similar with comparisons. Douglas and Nijssen (2003) presented a similar consumer mindsets, the instruments or scales compelling argument that consumer ethnocentrism may be used across cultures with some degree of may be understood differently in smaller countries confidence in the findings. In essence, this assumption that are dependent on imports given the absence of is pure etic since it is believed that the cultures are domestic manufacturers. In sum, they are contesting: comparable. But if this assumption cannot be made, the scale’s dimensionality, the premise that the etic and emic analyses are essential. While we may ethnocentrism concept is understood universally by assume that New Zealand and Singapore are similar to consumers, and if the concept and scale function the USA regarding the ethnocentrism concept given similarly in so-called “smaller” countries. their development, etic vs. emic analyses provide Our study represents a systematic procedure to proof of this assumption. validate the cross-cultural validity of the consumer Prudent verification of etic and emic seems ethnocentrism construct and its CETSCALE scale. We necessary for other consumer behavior constructs chose this construct for our investigation of emic and and their measures prior to cross-cultural etic issues since it is well established in the literature as demonstrated by its popularity. Indeed, the comparisons. Admittedly, verifying these properties CETSCALE has been cited well over 1000 times since seems a burden; yet in the absence of such it was introduced in 1987. In our study of New diagnostic work, cross-cultural findings may be Zealand – a commodity and agrarian based economy tenuous or patently incorrect. For example, the and Singapore – a trading country that has a growing consumer decision-making styles instrument has services-oriented economy, the initial qualitative been used in numerous cross-cultural studies, but findings show that the consumer ethnocentrism there has been no investigation of the emic vs. etic concept was understood by consumers in both nations; considerations of the construct or the scale. Theory hence qualitatively speaking, the concept passed the development and validation of cultural impacts of etic test regarding its conceptual understanding by these psychological constructs is hindered if etic and consumers. The quantitative results answered all the emic analyses are not conducted. In sum, cross- research questions in a positive way, providing etic cultural research that operates on the premise that validity psychometrically. We found the CETSCALE scales developed in North America can be applied to possess the following: uni-dimensionality with high universally to other countries is fraught with serious reliabilities, structural equivalence, metric (or shortcomings. If these concepts and scales are to be measurement unit) equivalence, scalar equivalence and used to “unlock the mindsets” of consumers in other individual items exhibiting non-differential countries, preliminary etic and emic considerations functioning. These results collectively demonstrate cannot be overlooked. Cross-cultural researchers are that the CETSCALE is not narrowly relevant to just encouraged to be on guard. References 1. Albaum, G. and Baker, K. (2005). The Imposed Etic in Survey Research: Fact or Fiction. Proceedings of the Fourth International Business and Economy Conference. Honolulu, Hawaii. Retrieved April 24, 2012, from http://scholar.googleusercontent.com/scholar?q=cache:rbrOQNoDQccJ:scholar.google.com/&h1=en&as_sdt=0.50. 2. Bentler, P. (1990). “Comparative Fit Indexes in Structural Models”, Psychological Bulletin, pp. 238-246.

14 Innovative Marketing, Volume 10, Issue 1, 2014

3. Berry, J. (1980). “Introduction to Methodology”, in H. Triandis & J. Berry (ed.). Handbook of Cross-Cultural Psychology (Vol. 2), Allyn and Bacon, Inc., Boston, MA. 4. Brown, T. (2006). Confirmatory Factor Analysis for Applied Research, The Guilford Press, New York, NY. 5. Craig C., and Douglas, S. (2005). International Marketing Research, Wiley & Sons, West Sussex, England. 6. de Mooij, M. (2010), Global Marketing and Advertising (3 ed.), Sage Publications, Thousand Oaks, CA. 7. de Ruyter, K., Van Birgelen, M., and Wetzels, M. (1998). “Consumer Ethnocentrism in International Services Marketing”, International Business Review, 7, pp. 185-202. 8. Douglas, S. and Craig, C. (1983), International Marketing Research, Prentice-Hall, Englewood Cliffs, NJ. 9. Douglas, S. and Nijssen, E. (2003). On the Use of Borrowed Scales in Cross-National Research: A Cautionary Note, International Marketing Review, 20 (6), pp. 621-642. 10. Durvasula, S., Andrews, J.C. and Netemeyer, R.G. (1997). “A Cross-Cultural Comparison of Consumer Ethnocentrism in the United States and Russia”, Journal of International Consumer Marketing, 9 (4), pp. 73-93. 11. Durvasula, S. and Lysonski, S. (2009). “How Offshore Outsourcing is Perceived: Why do Some Consumers Feel Threatened”, Journal of International Consumer Marketing, 21(1), pp. 17-33. 12. Henrich, J. and Boyd, R. (1998). “The Evolution of Conformist Transmission and the Emergence of Between- Group Differences”, Evolution and Human Behavior, 19 (4), pp. 215-241. 13. Henrich, J., Heine, S. and Norenzayan, A. (2010). “The Weirdest People in the World?” Behavioral and Brain Sciences, 33, pp. 61-135. 14. Herche, J., Swenson, M., and Verbeke, W. (1996). Personal Selling Constructs and Measures: Emic versus Etic Approaches to Cross-National Research”, European Journal of Marketing, 30 (7), pp. 83-97. 15. Kaynak, E. and Kara, A. (1996). “Ethnocentric Behavior of Consumers in Emerging Market Economy”, in AMA Winter Educators’ Confeence: Marketing Theory and Application, 7, American Marketing Association, Chicago, pp. 39-45. 16. Klein, J., Ettenson, R. and Krishnan, B. (2006). “Extending the Construct of Consumer Ethnocentrism: When Foreign Products are Preferred,” International Marketing Review, 23 (3), pp. 304-321. 17. LeVine, R., and Campbell, D. (1972). Ethnocentrism: Theories of Conflict, Ethnic Attitudes and Group Behavior. Wiley & Sons, New York, NY. 18. Lorenzo-Seva, U. and Ferrando, P. (2003). “IMINCE: An Unrestricted Factor Analysis Based Program for Assessing Measurement Invariance”, Behavioral Research Methods, Instruments, & Computers, 35, pp. 318-321. 19. Luque-Martinez, T., Ibanez-Zapata, J. and del Barrio-Garcia, S. (2000). “Consumer Ethnocentrism Measurement – An Assessment of the Reliability and Validity of the CETSCALE in Spain”, European Journal of Marketing, 34 (11/12), pp. 1353-1374. 20. Marcoux, J., Filiatrault, P. and Cheron, E. (1997). “The Attitude Underlying Preferences of Young Urban Educated Polish Consumers Towards Products Made in Western Countries”, Journal of International Consumer Marketing, 9 (4), pp. 5-29. 21. Marsh, H. (1994). “Confirmatory Factor Analysis Models of Factorial Invariance: A Multifaceted Approach”, Journal of Structural Equations Modeling, 1, pp. 5-34. 22. Mavondo, F. and Tan, A. (1999). “Reconceptualising the CETSCALE”, Proceedings of ANZMAC Conference. Retrieved May 21, 2006, from http://smib.vuw.ac.nz:8081/www/ANZMAC1999/Site/M/Mavondo_Tan.pdf. 23. Netemeyer, R., Durvasula, S. and Lichtenstein, D. (1991). “Assessing the Cross-National Applicability of the Consumer Ethnocentrism Scale,” Journal of Marketing Research, August, 320-328. 24. Nijssen, E., Douglas, S. and Bressers, P. (1999). “Attitudes Toward the Purchase of Foreign Products: Extending the Model,” Retrieved June 1, 2008, from http://pages.stern.nyu.edu/~sdouglas/rpubs/attitudes.html. 25. Reardon, J., Miller, C., Vida, I. and Kim, I. (2005). “The Effects of Ethnocentrism and Economic Development on the Formation of Brand and Ad Attitudes in Transitional Economies,” European Journal of Marketing, 39 (7/8), pp. 23-38. 26. Segall, M., Dasen, P., Berry, J. and Poortinga, Y. (1990). Human Behavior in Global Perspective: An Introduction to Cross-cultural Psychology, Pergamon, New York: NY. 27. Sekaran, U. (1983). “Methodological and Theoretical Issues and Advancements in Cross-Cultural Research”, Journal of International Business Studies, 14 (Fall), pp. 61-73. 28. Sharma, S., Shimp, T. and Shin, J. (1995). “Consumer Ethnocentrism: A Test of Antecedents and Moderators”, Journal of Academy of Marketing Science, 23 (1), pp. 26-37. 29. Shimp, T. and Sharma, S. (1987). “Consumer Ethnocentrism: Construction and Validation of the CETSCALE”, Journal of Marketing Research, 24 (August), pp. 280-289. 30. Steenkamp, J. and Baumgartner, H. (1998). “Assessing Measurement Invariance in Cross-National Consumer Research,” Journal of Consumer Research, 25, pp. 78-90. 31. Sumner, W. (1906). Folkways: The Sociological Importance of Usages, Manners, Customs, and Moral, Ginn & Company, New York, NY. 32. Suphellen, M. and Rittenburg, T. (2001). “Consumer Ethnocentrism When Foreign Products are Better”, Psychology & Marketing, 18 (9), pp. 907-927. 33. Tucker, L. (1951). A Method for Synthesis of Factor Analysis Studies, Department of the Army, Washington, D.C. 34. Van de Vijver, F. and Leung, K. (1997). Methods and Data Analysis for Cross-Cultural Research. Sage, Thousand Oaks, CA.

15 Innovative Marketing, Volume 10, Issue 1, 2014

35. Van de Vijver, F. and Leung, K. (2011). “Equivalence and Bias”, in D. Matsumoto, and F. Van de Vijver (ed.), Cross-Cultural Research Methods in Psychology, Cambridge University Press, New York, NY, pp. 17-45. 36. Van de Vijver, F. and Poortinga, Y. (1994). “Methological Issues in Cross-Cultural Studies on Parental Rearing Behavior and Psychopathology”, in Perry, C., Arrindell, W. and Eisemann, M. (ed.), Parental Rearing and Psychopathology, Wiley & Sons, Chicester, pp. 173-197. 37. Van de Vijver, F., Valchev, V. and Sunnet, I. (2009). Structural Equivalence and Differential Item Functioning in the Social Axioms Survey. In K. Leung, & M. Bond, Advances in Research on Social Axioms. New York, NY: Springer, pp. 51-80. 38. Vida, I. and Damjan, J. (2001). “The Role of Consumer Characteristics and Attitudes in Purchase Behavior of Domestic versus Foreign Made Products: The Case of Slovenia”, Journal of East-West Business, 6 (3), pp. 111-133. 39. Watters, Ethan (2013). “We Aren’t the World”, Pacific Standard, March/April, pp. 46-53.